Emergent BEC mechanism in flat-band superconductors

Abstract: The formation of bound bosonic pairs of fermions, followed by their Bose-Einstein (quasi) condensation (BEC), is a foundational mechanism of superconductivity. At low filling, flat-band superconductivity is well captured by this mechanism provided the flat band is separated from the occupied lower band by an energy gap. However, particularly high $T_c$ values are anticipated when the non-interacting flat and lower bands touch---as in the prototypical attractive Lieb-lattice model studied here---invalidating the conventional picture: while interactions might protect the bound state by opening a gap, no small parameter guarantees the separation of the bound-state energy from this gap or occupied-band excitations, leaving the pair's fate uncertain. Based on a controlled-precision numerical protocol---which we demonstrate to be essential in this fundamentally non-perturbative problem---we show that the BEC mechanism, underpinned by an interaction-induced gap, is generically robust and remarkably efficient: fermions doped into the flat band form bound pairs within this gap with an anomalously light effective mass, enabling an exceptionally high $T_c$.

• The paper demonstrates that Hartree-induced sublattice shifts open a gap of 0.875t in the gapless Lieb lattice at U=-2t, stabilizing the preformed-pair branch beyond mean field.
• Diagrammatic calculations and Monte Carlo confirm stable charge-2e pairs, whose correlation-renormalized effective mass is approximately 3.3/(t a₀²), about half the static-band estimate.
• The resulting light pairs support an exceptionally high dilute-limit BKT transition through Tc ≈ 0.65(nf/m*), although the full density dependence and behavior beyond U/t=-2 remain unresolved.

Setting and motivation

The paper addresses a long-standing ambiguity in the theory of flat-band superconductivity: whether the preformed-pair (bosonic) mechanism survives when the non-interacting flat band touches the lower occupied band, as in the standard attractive Hubbard model on the Lieb lattice. In gapped flat-band systems, the two-body problem in the empty-band subspace yields a well-defined bound state with finite effective mass $m^*$, and the transition temperature follows the universal dilute-Bose-gas relation $T_c \approx 0.65\,(n_f/m^*)$ (2608.19044). Without a gap, the bound-state energy per particle overlaps the occupied band, and no small parameter ($|U|/\Delta \ll 1$) guarantees stability of the pair branch. The paper resolves this for the gapless Lieb lattice at $U=-2t$ and filling $n=2$.

Interaction-induced gap

The central analytic observation is that Hartree mean-field (HMF) sublattice shifts, $E_a = U n_a/2$, open a gap $\Delta$ below the flat band once the lower band is filled. Self-consistent HMF at $U=-2t$ gives $n_1=1.250$, $n_2=n_3=0.375$, and $T_c \approx 0.65\,(n_f/m^*)$0. Because this gap is generated by the interaction itself, the problem is fundamentally non-perturbative, and the authors verify gap stability beyond mean field using a hierarchy of diagrammatic methods: the self-consistent one-loop Bold4 scheme, Bold4+ (exact to order $T_c \approx 0.65\,(n_f/m^*)$1 in the single-particle channel), and numerically exact diagrammatic Monte Carlo with combinatorial summation (DiagMC-CoS), with Dlog Padé and integral-approximant resummation and intrinsic error control. Vertex corrections partially reverse the screening-induced gap reduction, and the converged gap settles between the HMF and Bold4 values. The second band becomes weakly dispersive (bandwidth $T_c \approx 0.65\,(n_f/m^*)$2), so "flat band" is retained only nominally.

Bound pairs and effective mass

Two complementary protocols establish the existence of a stable charge-$T_c \approx 0.65\,(n_f/m^*)$3 quasiparticle. First, a variational two-body ansatz on the HMF-renormalized bands yields a bound state lying inside the gap ($T_c \approx 0.65\,(n_f/m^*)$4) with $T_c \approx 0.65\,(n_f/m^*)$5, i.e. $T_c \approx 0.65\,(n_f/m^*)$6. Second, grand-canonical diagrammatic calculations of the pair propagator, with the chemical potential placed between $T_c \approx 0.65\,(n_f/m^*)$7 and $T_c \approx 0.65\,(n_f/m^*)$8, extract $T_c \approx 0.65\,(n_f/m^*)$9 from the asymptotic decay of $|U|/\Delta \ll 1$0.

The stability margin $|U|/\Delta \ll 1$1 is identified as the most significant many-body quantity: in Bold4, screening from the occupied band drives $|U|/\Delta \ll 1$2 down to only a few percent of $|U|/\Delta \ll 1$3, placing the system near an instability and confirming the non-perturbative character of the problem. Vertex corrections in Bold4+ and DiagMC-CoS restore a comfortable positive $|U|/\Delta \ll 1$4, removing doubt about the stability of the interaction-induced band insulator. By contrast, $|U|/\Delta \ll 1$5 itself varies by only about 10% across schemes, while $|U|/\Delta \ll 1$6 varies by roughly 100% between HMF and diagrammatic treatments — a strong, quantitative demonstration that pair mobility is governed by correlation effects beyond the static band renormalization. Bold4 already captures the dominant mechanism (hole-particle excitation and recombination processes providing additional pair-propagation channels), and the Bold4+ and DiagMC-CoS results agree within error bars, giving a converged $|U|/\Delta \ll 1$7 — roughly half the variational estimate.

Consequences for $|U|/\Delta \ll 1$8

The anomalously light pair mass directly implies an exceptionally high BKT transition temperature in the dilute regime via $|U|/\Delta \ll 1$9. Combined with the previously established crossover scale $U=-2t$0 at half-filling, the paper sketches two candidate density dependences of $U=-2t$1: a non-monotonic peak (case A) or rapid saturation (case B). The authors do not resolve which obtains, but argue that the high critical temperatures found at $U=-2t$2 remain robust across a broad density range in either scenario.

Limitations and open questions

The conclusions rest on several qualifications the authors state explicitly. The variational two-body treatment excludes particle-hole excitations by construction, and the grand-canonical protocol requires exponentially dilute pairs, so the pair picture is validated only at low doping. The gap and binding-energy stability are demonstrated at the single coupling $U=-2t$3; behavior at other couplings, and the full $U=-2t$4 curve between the dilute-BKT limit and half-filling, remain open. The DiagMC-CoS simulations are performed at a higher temperature ($U=-2t$5) than the Bold-family calculations, and the extrapolation protocol relies on resummation assumptions whose residual uncertainty is folded into the error bars.

Conclusion

The paper shows, with controlled numerical precision, that the preformed-pair BEC mechanism is generically robust in gapless flat-band superconductors: the attractive interaction itself opens the single-particle gap that stabilizes the pair branch, and correlations beyond mean field produce a pair roughly twice lighter than a static-band treatment predicts. The result converts a formally ill-defined two-body problem into a quantitatively controlled dilute-Bose-gas description and pins down the low-density slope of $U=-2t$6 for the Lieb-lattice model.